In the design of gel phase change composite wall materials for solar greenhouses, the alteration of material composition could directly affect the thermal performance of gel phase change composite wall materials. In order to obtain better suitable gel phasechange composite wall material for solar greenhouses, Na₂HPO₄·12H₂O-based gel phasechange materials with different content of ingredient (Na₂SiO₃·9H₂O, C₃₅H₄₉O₂₉, KCl, and nano-α-Fe₂O₃) were obtained via the Taguchi method and machine learning algorithms, such as Support Vector Regression (SVR), Random Forest (RF), and Gradient Boosting Trees (GBDT). The result shows that the GBDT is more suitable for the thermal performance optimization prediction of gel phase change composite wall materials, including time cooling (TC), latent heat of phase change (ΔH_m), supercooling degree (ΔT), and phase change temperature (T_m). The determination coefficient (R²) of time cooling (TC), latent heat of phase change (ΔH_m), supercooling degree (ΔT), and phase change temperature (T_m) by GBDT are 0.9987, 0.99965, 1, and 0.9995, respectively. The mean absolute error (MAE) coefficient percentage of supercooling degree (ΔT), phase change temperature (T_m), latent heat of phase change (ΔH_m), and time of cooling (TC) by GBDT are 0.32%, 0.25%, 0.17%, and 0.26%, respectively. The root mean square error (RMSE) of supercooling degree (ΔT), phase change temperature (T_m), latent heat of phase change (ΔH_m), and time of cooling (TC) by GBDT are 0.41%, 0.32%, 0.19%, and 0.35%, respectively. The optimal result predicted by GBDT is Na₂HPO₄·12H₂O + 5% Na₂SiO₃·9H₂O + 12% KCl + 0.2% Nano-α-Fe₂O₃ + 3% C₃₅H₄₉O₂₉, which was verified by experiments.

The content of modified materials in multicomponent gel phase change materials directly affects their performance characteristics. To investigate the influence of different contents of modified materials on the performance features of Na₂HPO₄·12H₂O-based multicomponent Gel Phase Change Materials, four single factors (Na₂SiO₃·9H₂O, C₃₅H₄₉O₂₉, KCl, and nano-α-Fe₂O₃) and their interactions were selected as influencing factors. Using the Taguchi method with an L₂₇(3¹³) orthogonal array, multi-step melt–blending experiments were conducted to prepare a novel multi-component phase change material. The characteristics of the new multi-component phase change material, including supercooling degree (ΔT), phase change temperature (T_m), latent heat of phase change (ΔH_m), and cooling time (CT), were obtained. In addition, characterization techniques such as DSC, SEM, FT-IR, and XRD were employed to analyze its thermal properties, microscopic morphology, chemical stability, and crystal structure. Based on the experimental results, the signal-to-noise ratio (S/N) was used to rank the influence of each factor on the quality characteristics, and the p-value from analysis of variance (ANOVA) was employed to evaluate the significance of each factor on the performance characteristics. Then, the effects of each significant factor on the characteristics of the multiple gel phase change materials were analyzed in detail, and the optimal mixing ratio of the new multiple gel phase change materials was selected. The results showed that Na₂SiO₃·9H₂O, KCl, and α-Fe₂O₃ were the most critical process parameters. This research work enriches the selection of composite gel phase change materials for solar greenhouses and provides guidance for the selection of different modified material contents using Na₂HPO₄·12H₂O as the starting material.

In order to achieve fast and accurate life damage online monitoring of the steam turbine rotor, it was significant to propose an empirical-statistical model using a machine learning algorithm instead of finite element simulation to improve the effect of operation. The finite element method was used to calculate the maximum stress during the start-up schedule. The linear CDM (Continuum Damage Mechanics) and nonlinear CDM were applied to assess the creep-fatigue damage of the steam turbine rotor. A empirical-statistical model of a 600 MW steam turbine rotor was established by using temperature change rate and maximum stress according to the finite element result samples, which is proposed by compared R2 of SVR (Support Vector Regression), LSTM (Long Short-Term Memory) and RRM (Ridge Regression Method), which was also verified by finite element simulation under a random start-up parameters. The results showed that the creep-fatigue damage could be calculated by nonlinear CDM for more safety rather than linear CDM. The R2 of SVR (Support Vector Regression), LSTM (Long Short-Term Memory) and RRM were 0.9377, 0.9647 and 0.999, respectively. RRM was more suitable for the empirical-statistical model establishment of the steam turbine rotor. By comparing the empirical-statistical model result and finite element result under random parameters of the start-up schedule, the error is 0.51%.

Solar greenhouses are indispensable infrastructure for winter vegetable production in high-latitude regions, maintaining favorable growing conditions without auxiliary heating through superior thermal insulation and solar radiation capture. However, extreme meteorological events disrupt this equilibrium, and soil temperature prediction remains challenging due to thermal hysteresis—time-lagged responses to environmental drivers that threaten crop viability during pre-dawn periods. Existing studies focus on air temperature or short-term horizons (3–6 h), leaving the critical 12 h prediction window inadequately addressed. This study develops a hybrid machine learning framework for 12 h soil temperature prediction in a solar greenhouse. High-resolution data were collected using soil temperature sensors at 200 mm depth (tomato root zone) during the winter growing season. Five algorithms were evaluated: RF and XGBoost (Bagging/Boosting for trend/residual capture), LSTM and GRU (temporal memory), and a novel RF-XGBoost hybrid with two-stage residual correction. The hybrid model achieved R² = 0.9927, improving standalone RF (R² = 0.9883) by 0.45% with MSE, RMSE, MAE reductions of 18.1%, 9.6%, 13.9%. Maximum 12 h error was 2.52 °C versus 3.12–4.60 °C for standalone models. The 12 h horizon enables preemptive heating activation, mitigating frost risk while avoiding unnecessary energy expenditure.